Predicting Older People’s Thermal Sensation by a New Integrated Physiological-based and Data-driven Model

Due to age-related physiological changes, older people are more vulnerable than young people to heat or cold conditions. Predicting older people's thermal sensations is essential for controlling the built environment and avoiding extreme heat/cold injuries. Previous studies mainly focused on predicting the thermal sensation of young people, and the data-driven methods are often not constrained by physiological responses. This study proposes a new integrated model to combine the two-node physiological model and the data-driven method random forest classifier. The surveyed data of older people come from ASHRAE Global Thermal Comfort Database II. The dataSET has collected the environmental conditions, subjects' factors, and survey results of thermal sensation vote (TSV). In this study, with the environmental conditions (air temperature, mean radiant temperature, relative humidity, and airspeed) and subject factors (clothing insulation, height, and weight) as inputs, core and skin temperatures, water loss, and standard effective temperature (SET) can be calculated by the two-node model of older people. The above physiological parameters and building operation mode (naturally-ventilated/air-conditioned - NV/AC), older people's gender, surveyed seasons, and climate zones are used to train the data-driven model. The results show that the overall accuracy classification score of the integrated model is 90%, which is more accurate than the PMV model and the majority of other data-driven studies. The integrated model can also reach above 80% accuracy classification score under different building operation modes (NV/AC), older people's gender, surveyed seasons, and climate zones. The correlation between SET and TSV is better than the traditional linear regression. It is found that there is a possibility that older people's core temperature increase to a dangerous level (>38℃) even when they just feel slightly warm.